Mineral identification using indoor hyperspectral Bardia ... bardia/files/DSS- ¢ ...

download Mineral identification using indoor hyperspectral Bardia ... bardia/files/DSS- ¢  Fran£§ois

of 29

  • date post

    25-Aug-2020
  • Category

    Documents

  • view

    2
  • download

    0

Embed Size (px)

Transcript of Mineral identification using indoor hyperspectral Bardia ... bardia/files/DSS- ¢ ...

  • Bardia Yousefi, Saeed Sojasi, Clemente

    Ibarra-Castanedo, Georges Beaudoin ,

    François Huot, Xavier P. V. Maldague

    Martin Chamberland, and Erik Lalonde

    Mineral identification using indoor hyperspectral

    imaging

    Département de génie électrique et de génie informatique

    TELOPS, Inc., 100-2600 St-Jean-Baptiste Ave, Quebec, Qc, G2E 6J5, Canada

    Canada Research Chair in Multipolar Infrared Vision – Vision Infrarouge Multipolaire: MiViM

  • Outline

    • Introduction

    • Hyperspectral Imaging

    • Problem Statement

    • What we are doing

    • Experimental Results

    • Measurement

    • Method

    • HYMID

    • conclusion

  • HYPERSPECTRAL IMAGING

     Hyperspectral imaging, like other spectral imaging,

    collects and processes information from across the

    electromagnetic spectrum. The goal of hyperspectral

    imaging is to obtain the spectrum for each pixel in

    the image of a scene, with the purpose of finding objects,

    identifying materials, or detecting processes.

    3

    photonics.com

    Spatial

    Spectral

    Applications

    • vegetation

    • urban area

    • geology

    • target detection

    • medical

    • etc

    northropgrumman.com View of the Hyperspectral Imager for Coastal Oceans (HICO)

    nasa.gov

  • WHAT WE ARE DOING

    4

    Hyperspectral in Satellite

    Hyperspectral in Laboratory

  • INTRODUCTION(Problem Statement)

     Mineral identification is challenging in the field of geology

    and mineralogy.

     It relates to geological research and is usually conducted by

    geologists (mineralogy experts).

     It is extensively investigated by hyperspectral remote sensing

    (and airborne) sensing and has been the subject of many

    research studies.

     Manual identification of mineral samples by a mineralogy

    expert which is a time consuming process and provides high

    level of disparity due to fatigue or inadequate methods for this

    specific application.

    5

  • INTRODUCTION

     The main objective of this research lies under the

    automatic mineral identification which is considered

    as hyperspectral unmixing.

    Making automated system for mineral

    identification

     Hyperspectral unmixing in the hyperspectral

    infrared image.

     It use(d/s) for long time and most of the HSI

    researches are basically for remote sensing(and

    airborne) research. (e.g. Only in 2015 the geological based hyperspectral infrared imagery (contributions) papers

    were more 100> )

    What

    6

    Photograph showing measurements being made with the PIMA

    II field spectrometer. Picture from: Kruse, F. A. (1996). Identification and mapping of minerals in drill core using hyperspectral

    image analysis of infrared reflectance spectra. International journal of

    remote sensing, 17(9), 1623-1632.

  • Mineral identification

    7

    Data-mining

    Measurement

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    •Data retrieval

    Data-mining • Data retrieval

    • Creating a system for automatic

    mineral identification:

    •Feature extraction /Selection

    •Classification- Pattern

    recognition task & machine

    learning

  • 8

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    Measurement

  • 9

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    New mineral grains have been prepared for the new experiment

  • Hyperspectral imaging

    10

    HyperCam

    Heating Element

    Mineral Samples

    InfraGold

    Heating Element

    Imaging are done at Telops Inc. 02 April 2015

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

  • List of all samples

    11

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    1. ILM

    2. ILM – NEW

    3. OL

    4. OL – NEW

    5. CHR

    6. CHR – NEW

    7. PYR

    8. QTZ 1&2

    9. QTZ-NEW

    10. QTZ- FELD

    11. MIX 1

    12. MIX 2

  • 50

    100

    150

    200

    250

    20

    40

    60

    80

    100

    120

    140

    160

    180

    200

    305

    310

    315

    320

    325

    IR View

    IR ViewIR View

    IR View

    297.5

    298

    298.5

    299

    299.5

    300

    300.5

    301

    301.5

    302

    IR View

    297.5

    298

    298.5

    299

    299.5

    300

    300.5

    301

    301.5

    302

    Olivine - OL

    12

    50100150200250

    20

    40

    60

    80

    100

    120

    140

    160

    180

    200

    305

    310

    315

    320

    325

    10.77 mm

    FileName = '20150402_211927875_OL_ON.radiance.sc';

    OL 1

    OL2

    OL 3

    OFF (before) OFF (after)ON

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    ASTER

    OUR Test

  • Chromite - NEW- CHR

    13

    CHR1

    CHR2 CHR3

    Binocular

    Heating source OFFHeating source ON

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    ASTER

    OUR Test

  • ASTER - Chromite

    14

    Current Band

    Possible Signature

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    Picture adopted from ASTER

  • ASTER – Ilmenite & Chromite

    15

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    Current Band

    Possible Signature

    Current Band

    Possible Signature

    There is no signature for Ilmenite and Chromite in LWIR

    Pictures are adopted from ASTER

  • Ilmenite – ILM- NEW

    16

    Binocular

    Heating source OFFHeating source ON

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    OUR Test

  • Pyrope - PYR

    17

    PYR1

    PYR2 PYR3

    Binocular

    Heating source OFFHeating source ON

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    ASTER

    OUR Test

  • QUARTZ –NEW - QTZ

    18

    QTZ1

    QTZ2QTZ3

    Binocular

    Heating source OFF Heating source ON

    Measurement •Mineral preparation

    •Conducting the experiments

    •Producing the clean data

    •Finding the reference datasets

    ASTER

    OUR Test

  • HYMID Hyperspectral Mineral Identifier (HYMID): a tool for applying spectral analysis

  • Reference (EEA)

    ASTER/JPL

    Approx. EE

    METHODOLOGY

    Matched

    Filter

    Spectral Angle

    Mapper

    Abundance

    Map

    MAM map

    Material abundance map (MAM) generation

    Spectral Comparison

    Input HSI

    Normalize

    cross

    correlation

    NormXCorr

    SAM

    MF

    SID

    Supervised

    ELM

    SVM

    Unsupervised

    Kernelled HSV K-

    mean

    Spectral Info.

    Divergence

    Matched

    Filter

    Spectral Angle

    Mapper

    Normalize

    cross

    correlation

    NormXCorr

    SAM

    MF

    SID

    Spectral Info.

    Divergence

    First Derivative

    (FD)

    Matched

    Filter

    Spectral Angle

    Mapper

    Normalize

    cross

    correlation

    NormXCorr

    SAM

    MF

    SID

    Spectral Info.

    Divergence

    Signature band

    (SFF)

  • RESULTS

    21

    Simple looking at the problem, applying a simple threshold in the certain wavelength.

    Feng, J., Rivard, B., Rogge, D., & Sánchez-Azofeifa, A. (2013). The longwave infrared (3–14μm) spectral properties of rock encrusting lichens based on laboratory spectra and

    airborne SEBASS imagery. Remote Sensing of Environment, 131, 173-181.

  • RESULTS

    22